Pengxiang Meng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Pengxiang Meng has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Xiaowen Tao
School of Computer Science and Statistics, Trinity College Dublin, Dublin, Ireland
2 shared publications - Bing Zhu
Jilin University
1 shared publication - Bing Zhu
National Key Laboratory of Automotive Chassis Integration and Bionics and the College of Automotive Engineering, Jilin University, Changchun, China
1 shared publication Bo Wang
1 shared publication- Dongjian Song
State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun, China
1 shared publication - Hongyi Jiang
National Key Laboratory of Automotive Chassis Integration and Bionics and the College of Automotive Engineering, Jilin University, Changchun, China
1 shared publication
Latest Publications
- 11 August 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 333
DiffAMF: A diffusion-based adaptive multimodal fusion model for driver attention prediction
- 11 August 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 333
DiffAMF: A diffusion-based adaptive multimodal fusion model for driver attention prediction
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 12 (2) , 3280-3291
MVP: A Multisensor Vehicle-Mounted Positioning Framework for Autonomous Driving
Positioning technology plays a critical role in environmental perception and serves as the foundation for decision-making and motion control in autonomous vehicles. Existing positioning frameworks often face challenges such as limited accuracy, high costs, or poor adaptability to diverse environments. To address these issues, this article proposes a multisensor vehicle-mounted positioning (MVP) framework for autonomous driving, leveraging the complementary strengths of multiple sensors and advanced graph optimization techniques. The proposed MVP framework integrates data from inertial measurement units (IMUs), cameras, and global positioning systems (GPSs) to achieve robust and precise positioning. Through nonlinear optimization, the system tightly integrates vision-based odometry with IMU data to enhance short-term motion estimation, even in challenging scenarios such as feature-sparse environments and abrupt maneuvers. Furthermore, the incorporation of a novel outlier rejection mechanism significantly boosts positioning accuracy and adaptability, ensuring robust performance in complex and dynamic conditions. To ensure global positioning consistency, the framework employs adaptive fusion of visual-inertial odometry (VIO) and GPS data, enhanced by real-time kinematic (RTK) processing, through global pose graph optimization. This approach effectively balances positioning accuracy and cost while enhancing the system’s adaptability to diverse operating environments. Experimental validation shows promising results with 0.7359 m mean absolute pose error (APE) on the KITTI dataset (79.60% improvement over ORB-SLAM) and 0.20-m accuracy in real-world campus testing. Results demonstrate that the MVP framework outperforms existing approaches in terms of accuracy, stability, and cost-effectiveness, making it a promising solution for real-world autonomous driving applications such as campus shuttles, depot logistics, and low-speed urban operation domains.
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 12 (2) , 3280-3291
MVP: A Multisensor Vehicle-Mounted Positioning Framework for Autonomous Driving
Positioning technology plays a critical role in environmental perception and serves as the foundation for decision-making and motion control in autonomous vehicles. Existing positioning frameworks often face challenges such as limited accuracy, high costs, or poor adaptability to diverse environments. To address these issues, this article proposes a multisensor vehicle-mounted positioning (MVP) framework for autonomous driving, leveraging the complementary strengths of multiple sensors and advanced graph optimization techniques. The proposed MVP framework integrates data from inertial measurement units (IMUs), cameras, and global positioning systems (GPSs) to achieve robust and precise positioning. Through nonlinear optimization, the system tightly integrates vision-based odometry with IMU data to enhance short-term motion estimation, even in challenging scenarios such as feature-sparse environments and abrupt maneuvers. Furthermore, the incorporation of a novel outlier rejection mechanism significantly boosts positioning accuracy and adaptability, ensuring robust performance in complex and dynamic conditions. To ensure global positioning consistency, the framework employs adaptive fusion of visual-inertial odometry (VIO) and GPS data, enhanced by real-time kinematic (RTK) processing, through global pose graph optimization. This approach effectively balances positioning accuracy and cost while enhancing the system’s adaptability to diverse operating environments. Experimental validation shows promising results with 0.7359 m mean absolute pose error (APE) on the KITTI dataset (79.60% improvement over ORB-SLAM) and 0.20-m accuracy in real-world campus testing. Results demonstrate that the MVP framework outperforms existing approaches in terms of accuracy, stability, and cost-effectiveness, making it a promising solution for real-world autonomous driving applications such as campus shuttles, depot logistics, and low-speed urban operation domains.